DictatorFlow

Overview

DictatorFlow is a high-performance voice dictation engine available as a native desktop app for Mac, Windows, and Linux, and as a low-latency REST/WebSocket API. It achieves a 1.2% word error rate on LibriSpeech test-clean and 150ms time-to-first-token, outperforming OpenAI Whisper, Google Cloud STT, AWS Transcribe, Deepgram, and AssemblyAI. Written in Zig with native binaries for Apple Silicon and Intel — no Electron bloat. Audio is never stored on servers, with on-device privacy-preserving algorithms. Supports 99+ languages with automatic detection and translation. Run models entirely on your local GPU for full offline operation.

Pricing: Pro: $9/month | Lifetime: $99 one-time | API: $0.004/sec

Architecture: Apple Silicon, Intel

Key Features

• 1.2% word error rate on LibriSpeech test-clean benchmark
• 150ms time-to-first-token latency
• Native binary written in Zig: no Electron, no bloat
• Fully local/offline mode running models on your GPU
• Audio never stored on servers with privacy-preserving algorithms
• 99+ languages with automatic detection and translation
• REST and WebSocket API with multi-provider fallback chain
• Supports PCM, WAV, WebM, MP3, and OGG formats
• BYOK (Bring Your Own Provider Key) support
• Cross-platform: Mac (Apple Silicon & Intel), Windows, Linux

Tags

Similar Apps